How to Replicate Competitor UGC Videos with AI: A Complete Workflow Using Gemini 3 Pro and Video Generation Models

With the release of Gemini 3 Pro, the OpenRouter platform now supports video as an input—a breakthrough that dramatically changes how brands can analyze and recreate user-generated content (UGC). Even more exciting, AI can now accurately detect every shot, person, action, object, background element, and brand signal inside a video.

This technical leap unlocks something previously impossible:

Brands can now replicate competitor UGC content automatically using AI.

By letting AI interpret the original UGC video and then using your own brand and product context, you can generate high-quality look-alike videos—at scale, with consistency, and without needing an entire production crew.

---

Step 1: AI Breaks Down Competitor UGC into Shot-Level Structure

Let’s take an example UGC video from the children’s brand Lovevery:

👉 Video link:

When this video is passed into Gemini 3 Pro or another multimodal AI model, it produces a detailed, structured analysis of every shot.

Here’s an actual example (Shot 1):

{
  "audio": "Upbeat instrumental music begins",
  "style": "Realism, candid lifestyle",
  "action": "A woman in casual blue clothing opens the door, bends down to pick up the branded box from the doorstep, and carries it inside",
  "shot_id": 1,
  "subject": "Woman, Package",
  "end_time": "00:05",
  "start_time": "00:00",
  "composition": "Full body shot",
  "contains_brand": true,
  "contains_product": true,
  "duration_seconds": 5,
  "context_environment": "Urban street entrance, brick building facade",
  "first_frame_description": "A static wide shot captures the exterior entrance of a modern apartment building with a beige brick facade and a large glass door...",
  "ambiance_colour_lighting": "Natural daylight, neutral tones, soft contrast",
  "camera_motion_positioning": "Static wide shot at eye level"
}

The AI extracts everything needed to understand and recreate the video:

• shot timing
• actions
• composition
• lighting
• environment
• brand/product presence
• and even a detailed first-frame image description

This becomes the foundation for generating brand-new UGC content.

---

Step 2: Provide Your Brand & Product Context to Generate a “Replica Script”

Once the competitor UGC is broken down, the next step is to inject your brand’s context:

• product features
• target customer
• usage scenario
• branding tone/style

By combining this with Google’s Veo 3.1 prompting best practices, models like Gemini 3 Pro can produce a shot-by-shot script tailored to your brand—while keeping the pacing and structure of the original competitor video.

For example:

The AI now outputs a full set of instructions for generating an equivalent video—same style, same flow, different brand.

---

Step 3: Use “First-Frame Descriptions” to Generate Key Images for Each Shot

Inside each AI-generated JSON block is a first_frame_description. This description can be used directly as a prompt for an image generator such as Nano Banana, producing a visual reference for each shot.

Why this matters:

• Video models rely on strong starting frames
• Having consistent image style boosts realism
• You maintain stylistic continuity across all shots

After generating the images, you’ll have:

✔ a shot-level script ✔ a reference image for each shot ✔ brand-aligned context ✔ consistent visual direction

---

Step 4: Generate Video Clips and Merge Them to Produce the Final UGC Video

With scripts and images ready, the next step is to:

1. Use a video generation model (Veo, Runway, etc.) to create each video clip
2. Use a stitching/merge tool (e.g., FAL) to combine all generated segments

This produces a complete, polished UGC video for your brand.

The entire process can run automatically and repeatedly—perfect for brands needing consistent UGC output.

---

Full Workflow Summary

This workflow enables fast, repeatable, scalable UGC creation—powered entirely by AI.

---

Explore the Full Automated System in Flowtra AI

The entire competitor replication workflow has been productized inside Flowtra AI

Flowtra’s interface makes the entire process plug-and-play—even for non-technical users.